Epiphenomenal Scare: The Rise of ChatGPT and the Imperatives for DevOps (Software) Engineers

Artificial intelligence (AI) in some form has been with us since decades. The global AI market size was estimated at US$ 119.78 billion in 2022 and is expected to hit US$ 1,591.03 billion by 2030 with a registered CAGR of 38.1%. The recent advancement in the field of AI is largely attributed to the use of neural networks to solve real-world problems where a computer program tries to emulate the functioning of the human brain. For decades, organizations around the world and industries have worked in this direction where they can have AI systems that can comprehend human language and interact with them in a meaningful way, known as natural language processing (NLP). ChatGPT, which is developed by OpenAI, tries to solve this problem. ChatGPT is a large language model (LLM), based on the transformer architecture, which is suited to generate text based on some input from the user. Its success is owed to a training methodology known as reinforcement learning from human feedback. In a short period, it has been adopted by companies across size and sectors ex. Microsoft, Bain and Company, Shopify, Salesforce, Morgan Stanley, Air India, Koo, Khan Academy etc.

This disruptive AI technology helps in the area of translation, summarization, and conversation. According to OpenAI, ChatGPT was trained on 45 terabytes of data available on the internet in an unsupervised way and has 175 billion parameters. It took just five days to reach 1 million users and a staggering 100 million by January 2023. It jumped to 1 Billion in February 2023, 1.6 Billion in March 2023, which is a testament of its popularity. Its predecessors were GPT-1 – introduced in 2018 – and GPT-2, followed by GPT-3 in successive years. ChatGPT, built on top of GPT-3.5, was then launched in November 2022. Using a verbal-linguistic IQ test, ChatGPT has an IQ of 155 which is superior to 99.9 % of existing test-takers. ChatGPT has passed the United States bar exam, CPA exam and U.S. medical licensing exam, Wharton MBA Exam, Google Coding interview- to name a few. 

DevOps is the practice of leveraging the agile and lean methodology to develop software applications. The job of these engineers is to effectively communicate between the software developers and the engineers managing the IT operations. It is a holistic approach that combines practices, methodologies, and technical tools to improve collaboration, communication, and integration between the development teams and operations teams to reduce the ‘time to market’ of a software product. Its automation tools can help reduce human error, and save time and effort for DevOps teams. ChatOps is a contemporary development that facilitates an easier integration between teams and various tools/ platforms of DevOps for effective development, testing and support.

Use cases of ChatGPT for DevOps Professionals:

Data Pipeline automation: The data pipeline is a method to assemble information from different data sources and pass them to various applications or machine learning algorithms for further processing. It is a challenging task to identify the relevant datasets as organizations these days are generating a lot of data. ChatGPT can play an important role here.

Machine Learning:  ChatGPT is a machine learning model and is trained on a massive dataset, hence, other models can be built leveraging its exceptional capabilities. To achieve this objective, DevOps engineers can collaborate with data scientists to build faster and more efficient machine learning models. Kubiya Inc., a startup that poses itself as “ChatGPT for DevOps,” recently announced the official launch of the first AI assistant for engineering platforms and knowledge management.

Automated Testing: As ChatGPT is trained over a lot of codes, it can fix erroneous codes in a faster way. It can also refactor the code and can assist in visualisation.

Improving user experience: A successful organization needs to augment the experience of its products and services. But the advent of LLM can help the organizations to create better and customized chatbots with the help of ChatGPT.

Expansion to new DevOps tasks: ChatGPT can be trained to automate new DevOps tasks, such as compliance management, disaster recovery leading to cost optimization.

Integration with voice assistants: ChatGPT can be integrated with voice assistants such as Alexa and Cortana to provide real-time assistance and feedback for DevOps tasks.

Infrastructure management and monitoring: ChatGPT can help provisioning, configuring and providing real-time feedback on performance and availability.

Deployment automation: ChatGPT can help automate the deployment process by running tests, building artifacts, and deploying code across environments. It can also integrate with continuous integration/continuous delivery pipelines for consistency and reliability.

Incident management and troubleshooting: ChatGPT can help identify and diagnose incidents by analyzing logs and providing real-time alerts, feedback and troubleshooting. Capital One, a leading financial institution has automated its incident management process that reduced the time and effort required to resolve incidents and improve the overall reliability of their systems.

ChatGPT enhances the careers of DevOps professionals by enabling them to develop new skills. With routine and mundane tasks getting automated through ChatGPT, developers can take on more challenging roles leading to increased job satisfaction and faster career growth. It also facilitates in creation of online communities where developers can share knowledge, learn from others, and collaborate on projects. Through ChatGPT, young developers can learn the new programming languages and can update themselves with latest trends and techniques.

Imperatives for Technical Team and Organisation:

Despite its numerous advantages, ChatGPT poses many economic, social, technological, legal challenges. Due to these, organisations need to implement strict governance mechanism requiring continuous top management intervention. Regular audits and monitoring should be conducted to identify potential vulnerabilities. Culture of innovation and cross-functional collaboration must be fostered by the management among the employees. HR should also rethink its learning and development policies in line with contemporary AI developments.

As DevOps teams are overworked, rather than fearing the AI tools like ChatGPT, they should welcome assistance in offloading the mundane tasks to focus on strategic tasks. NVIDIA, a leading graphics processing unit manufacturer, uses ChatGPT to automate their code review process reducing the workload of their development team and improve the quality of their code. To stay relevant in the face of AI-driven automation, DevOps professionals must focus on upskilling and reskilling. They must participate in online communities and forums to exchange ideas, share knowledge, learn from peers’ experiences, and state-of-the art know how.

Conclusion:

ChatGPT is a powerful tool that can be defined in terms of code generator, bug solver, documentation generator, real-time translator, collaborator and knowledge disseminator. By leveraging the power of NLP, it can help reduce human error, improve communication and collaboration, and save time and efforts for tech teams. As this chatbot continues to evolve at a frenetic pace, its potential in DevOps (and other broad spheres) will only rise, making it indispensable for DevOps teams. Generative AI like ChatGPT changes the work routine for technology professionals exponentially, necessitating their expertise from coding, testing, deploying etc. to prompt engineering (leveraging the expertise of ChatGPT by prompting right questions).

With current capabilities, ChatGPT is not in a position to completely replace human talent which is full of knowledge, skills, creativity and imagination along with emotions, needed for delivering real value to the end-customers. Nonetheless, it should be adopted by the DevOps professionals as a supplementary tool for making the routine tasks faster. ChatGPT also has limitations in terms of being biased due to training data, lack of contextual knowledge while offering solution and limited creativity and imagination compared to human beings. It also raises many ethical concerns that compel the organisations across the sectors to evaluate its pros and cons before its large-scale adoption and usage.

-By Nityesh Bhatt, Omkar Sahoo, Professor – Institute of Management, Nirma University, Ahmedabad